[G33] xserver-xorg-video-intel: TMDS-2 not discovered on dual-DVI ADD2 card

[Greg Edwards]

Binary package hint: xserver-xorg-video-intel

The X intel driver in current intrepid (2:2.4.1-1ubuntu10.3) and in jaunty alpha 5 (2:2.6.1-1ubuntu3) does not correctly discover and initialize the 2nd DVI port on a dual-dvi ADD2 card. The card is a Wintec Pegasus ADD2 card:

http://www.wintecindustries.com/orderdesk/products/productInfo.aspx?category_id=2129&product_id=21327

Only TMDS-1 is detected and initialized. The issue is the same as reported in:

Bug 19770: Q965 dual-head DVI with ADD2 2nd panel can't be found
https://bugs.freedesktop.org/show_bug.cgi?id=19770

I applied a modified version of the patch in comment #22 to the 2:2.4.1-1ubuntu10.3 driver in intrepid, and it
resolved the issue fine. Note the issue also exists in jaunty alpha5.

[lspci]
00:00.0 Host bridge [0600]: Intel Corporation 82G33/G31/P35/P31 Express DRAM Controller [8086:29c0] (rev 02)
 Subsystem: Holco Enterprise Co, Ltd/Shuttle Computer Device [1297:3106]
00:02.0 VGA compatible controller [0300]: Intel Corporation 82G33/G31 Express Integrated Graphics Controller [8086:29c2] (rev 02)
 Subsystem: Holco Enterprise Co, Ltd/Shuttle Computer Device [1297:3106]

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22291)
novell hwinfo command run with --gfx

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22292)
novell hwinfo command run with --monitor

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22293)
X11 log file

This log file just shows that the one panel did come up. Some warnings/errors
are in the output.

However, Novell's X11 config file sax2 wasn't able to configure the dual-head
xinerama mode. This makes sense since 'xrandr' doesn't even show it. So there
is no surprise it wasn't configured.

I did have some hand-crafted Xorg.conf files that tried to use the 2nd monitor
like I did back in 10.3 but since it didn't work I won't bother including it.

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22294)
xorg.conf, as created by novell sax2 x11 config tool

[Erikj-sgi (erikj-sgi)]

This is what the Novell engineer said when closing the novell bug.
(https://bugzilla.novell.com/show_bug.cgi?id=469878):

   You're right. TMDS-2 output is no longer detected by the current intel driver.

   I'm afraid this driver issue needs to be addressed by Intel's Linux
   driver development team. Thus I would like to ask you to report this
   issue upstream on bugs.freedesktop.org according to

(snip).

My machine has multiple roots so I can reboot my work desktop to run commands
or tests from time to time. -Erik

[Erikj-sgi (erikj-sgi)]

I plan to add some more details required by
2. Information to provide:

Yet. Looks like I neglected some steps there and I'll fill in the gap.

[Erikj-sgi (erikj-sgi)]

Some additional detail in suggested format:

Bug description:
(Please see first comment in this bug)

System environment:
-- chipset: Q965
-- system architecture: x86_64
-- xf86-video-intel: libvgahw.so compiled for 1.5.2, module ver 0.1.0,
                     ABI class X.org Video Driver 4.1
-- xserver: 1.5.2
-- mesa: OpenGL version string: 1.4 Mesa 7.2
-- libdrm: 2.4.1
-- kernel: 2.6.27.7-9-default
-- Linux distribution: OpenSUSE 11.1
-- Machine or mobo model: Dell Optiplex 745
-- Display connector: DVI to mainboard, 2nd DVI to ADD2 card, VGA unused

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22295)
xrandr --verbose

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22296)
Xorg.0.log while ModeDebug set to yes

[Erikj-sgi (erikj-sgi)]

Created an attachment (id=22297)
xorg.conf showing ModeDebug set to yes for the xorg log just attached

[Erikj-sgi (erikj-sgi)]

I mis-spoke in the comments above (too many systems, too little time).

Optiplex 745 has integrated VGA connector but no DVI.
Both of my DVI connections are on the ADD2 card. Only one is seen.
I believe everything else I wrote is correct. Sorry about that.

[Ling-ma (ling-ma)]

Hi Erik
Could you try adding option ForceSDVODetect true in device section of
xorg.conf?

Thanks
Ma Ling

[Erikj-sgi (erikj-sgi)]

Sure; I'll do this on Monday as the machine with the issue is at the office.
Thanks - I look forward to trying it. -Erik

[Ling-ma (ling-ma)]

hi Erik,

You don't need to use option ForceSDVODetect.

I think the root cause is from SDVOC failed to detect in 2.5.0 version.
In our latest version, we can detect it normally. please update your driver.

Thanks
Ma Ling

[Sndirsch-suse (sndirsch-suse)]

Hmm. Just a wild guess. Is this the fix?

Author: Zhenyu Wang <email address hidden>
Date: Wed Dec 10 10:54:31 2008 +0800

    Try to always probe SDVOC on 965G/965GM

    Detect bit of SDVOC is reserved on 965G/965GM, instead of ignore SDVOC
    this trys to always probe it on these chipsets.

diff --git a/src/i830_driver.c b/src/i830_driver.c
index 4f87efb..7590257 100644
--- a/src/i830_driver.c
+++ b/src/i830_driver.c
@@ -919,7 +919,9 @@ I830SetupOutputs(ScrnInfoPtr pScrn)
            i830_hdmi_init(pScrn, SDVOB);
       }

- if ((INREG(SDVOC) & SDVO_DETECTED) || pI830->force_sdvo_detect) {
+ if ((INREG(SDVOC) & SDVO_DETECTED) || pI830->force_sdvo_detect ||
+ /* SDVOC detect bit is reserved on 965G/965GM */
+ (IS_I965G(pI830) && !IS_G4X(pI830))) {
         Bool found = i830_sdvo_init(pScrn, SDVOC);

         if (!found && SUPPORTS_INTEGRATED_HDMI(pI830))

[Ling-ma (ling-ma)]

Yes, it should work fine now!

thanks
Ma Ling

[Sndirsch-suse (sndirsch-suse)]

Eri, please try the option first. Thanks.

[Erikj-sgi (erikj-sgi)]

I'm happy to report that adding the option indeed allowed both TMDS device sto
be discovered and reported by xrandr.

I had mirrored output to both panels at that point.

I'm having some other (unrelated) issues getting Xinerama going but I'll
work through those hopefully.

[Erikj-sgi (erikj-sgi)]

I'm having trouble getting Xinerama dual-head working with my old xorg.conf
file from openesuse10.3 that used to work.

I tried to craft a opensuse11.1 style one and also had trouble.

However, the dynamic method using xrandr to position stuff is working fine
for me now. Therefore, I think this specific issue is addressed. Thanks
much.

Worst case if I can't get xorg.conf working right is I'll just use xrandr for
a while; I don't log out much anyway :) Thanks again for looking at this.

[Sndirsch-suse (sndirsch-suse)]

Thanks for testing, Erik. I'm going to apply the patch so you no longer need to add this option to xorg.conf.

[Ling-ma (ling-ma)]

hi Erik,

Thanks for your help, I close the issue now.

Regards
Ma Ling

[Ling-ma (ling-ma)]

Created an attachment (id=22930)
please try the patch on your machine, thanks.

hi Erik,

Because we have not similar HW, whould you please try our latest patch on your machine under the same environment with that of this bug #19770, then paset Xorg log file?

Thanks a lot!
Ma Ling

[Edwardsg (edwardsg)]

Created an attachment (id=23445)
Xorg.0.log with dual-dvi SDVO card

I can confirm the patch from comment #22 works fine on my G33 chipset system
with a dual-dvi SDVO card.

[Greg Edwards (gedwards)]

Binary package hint: xserver-xorg-video-intel

The X intel driver in current intrepid (2:2.4.1-1ubuntu10.3) and in jaunty alpha 5 (2:2.6.1-1ubuntu3) does not correctly discover and initialize the 2nd DVI port on a dual-dvi ADD2 card. The card is a Wintec Pegasus ADD2 card:

http://www.wintecindustries.com/orderdesk/products/productInfo.aspx?category_id=2129&product_id=21327

Only TMDS-1 is detected and initialized. The issue is the same as reported in:

Bug 19770: Q965 dual-head DVI with ADD2 2nd panel can't be found
https://bugs.freedesktop.org/show_bug.cgi?id=19770

I applied a modified version of the patch in comment #22 to the 2:2.4.1-1ubuntu10.3 driver in intrepid, and it
resolved the issue fine. Note the issue also exists in jaunty alpha5.

[Bryce Harrington (bryce)]

Hi gedwards,

Please attach the output of `lspci -vvnn`, and attach your /var/log/Xorg.0.log (and maybe Xorg.0.log.old) file from after reproducing this issue. If you've made any customizations to your /etc/X11/xorg.conf please attach that as well.

[This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

[Greg Edwards (gedwards)]

This isn't chipset specific. The same thing happens when I stick the card in a 965 chipset system.

[Greg Edwards (gedwards)]

I think we likely just need these two commits:

Fix SDVO/HDMI detect
http://cgit.freedesktop.org/xorg/driver/xf86-video-intel/commit/?id=9d464bd5b0d2724f5edb26e859888ceb6a248f9b

Safely init SDVO found variable
http://cgit.freedesktop.org/xorg/driver/xf86-video-intel/commit/?id=506bbb8341c052499057834a669b82787454b426

I'll try respinning the driver with these and test it this evening.

[Greg Edwards (gedwards)]

Confirmed these two patches resolve the issue. Respun against Intrepid 2:2.4.1-1ubuntu10.3 driver.

[Bryce Harrington (bryce)]

In a deeper review of this I'm not entirely confident that it will be regression-free so am going to hold off on them for now. Also, the patches don't apply cleanly against the -intel we have in Jaunty so would need to be rewritten for that.

[Greg Edwards (gedwards)]

This works fine in jaunty now (xserver-xorg-video-intel 2:2.6.3-0ubuntu9), so this can be closed as fixed in 9.04.

[Andrew Ash (ash211)]

Closed per OP's report of being fixed in Jaunty.

[ArgelBargel (karsten-kraus)]

It works, but I had to enforce the SVDO-detection in xorg.conf via

Option "ForceSDVODetect" "true"

to get dual-head working.

Output of sudo lspci -vnn:

00:02.0 VGA compatible controller [0300]: Intel Corporation 4 Series Chipset Integrated Graphics Controller [8086:2e12] (rev 03)
 Subsystem: Fujitsu Siemens Computers Device [1734:114c]
 Flags: bus master, fast devsel, latency 0, IRQ 2298
 Memory at fc000000 (64-bit, non-prefetchable) [size=4M]
 Memory at e0000000 (64-bit, prefetchable) [size=256M]
 I/O ports at 1c70 [size=8]
 Capabilities: [90] Message Signalled Interrupts: Mask- 64bit- Queue=0/0 Enable+
 Capabilities: [d0] Power Management version 2